Hand Gesture Recognition Based on Surface Electromyogram Signal (sEMG) with Muscular Contraction Level and Real Time Implementation on An Artificial Prosthetic Wrist Using Artificial Neural Network (ANN)

Abstract

Hand, the main grasping organ of the human body, is controlled by the brain activity during particular work. This paper discusses about determining the patterns of five particular hand gestures by using the raw surface electromyogram signal (sEMG) with muscular contraction level (flex sensor signals) produced during the flexion of the muscles. The Artificial Neural Network (ANN) is trained by calculating time domain features from both of the sEMG and flex sensor signals. Using the trained neural network, when new data is taken in real time an artificial prosthetic wrist replicates the same hand gestures. The contribution of the paper is the addition of flex signals with sEMG. It is found that the training and testing accuracy is better when both sEMG and flex signals are used for healthy person. Another finding is that, for disable person, an artificial prosthetic wrist can replicate the hand gesture using sEMG signal but no significant change has been found using flex signal during the gesture recognition process.